Every game on this site had to pass a hard test before we built it. First we write the rules. Then we make computer players fight each other with those rules, thousands of times, and we measure what decides the winner. Skill has to matter more than luck. Smart choices have to matter more than fast fingers. And no boring trick may beat everything else. Most ideas fail somewhere on that list. This is a tour of the graveyard — real ideas, with the real numbers that killed them — and what each one taught the games that made it.
Killed for being a reflex test
Our games are made for two people on one couch. So one rule matters most: the better thinker should win, not the person with younger thumbs. This rule kills more ideas than everything else combined. We once tried a dodgeball duel. The measurements showed that reacting 0.15 seconds late hurt you about five times more than playing badly did. So it was not really a dodgeball game. It was a reaction test in a dodgeball costume. Killed. Next came a king-of-the-hill duel. We rebuilt it five times, and every version turned into a race of tiny, twitchy moves. Killed. A bouncing-bullet duel came closest of all. But a player who thought 0.15 seconds late still lost 93% of the time. Every knob we turned to fix that also made the game duller. The two problems were one problem. Killed.
Those corpses shaped the games you can play now. In Wake, you move between three lanes, like changing lanes on a road. Choosing the right moment beats twitching fast. We measured it: slow reactions cost about 2 points of win rate there, while a better plan is worth ten to fifty.
Killed because the fun part never happened
One idea sounded wonderful. A rope-wrestling duel beside a pit: grab the rope, swing your opponent, throw them in. We built the rules four different ways. Across all of that testing, the throw worked exactly zero times. Why? The throw only worked when both players stood in just the right spots. But your opponent controls where they stand — and they never once agreed to lose. A cousin of this idea died the same way. Each player swung a heavy ball on a rope. The spinning ball turned out to be a shield, not a weapon, and the defensive spinner won 85% of its matches.
The lesson became a law for every game we ship: a move must work from anywhere, not just from one magic position. That is why our games are built around things both players chase — the puck in Neon Hockey, the ball in Spike!, the shove in Sumo Smash. Nobody can refuse a puck.
Killed by one boring trick
Some ideas die because one dull plan quietly beats everything. We tried a duel on thin ice, where the ice breaks behind you as you skate. The plan was to trap your opponent on a shrinking piece. The measurements found something sad instead. The best plan was to ignore your opponent completely and skate off across the fresh ice. The chaser lost every single match. Our exciting duel was really two people skating alone, side by side. Killed. Even Fuse — a game that shipped — had a version like this. In that version, the bomb scorched the tile it landed on. So the best move was to throw the bomb right at your opponent's feet, burning their favorite floor tile every turn. That friendly-looking move beat everything, about 65 matches out of 100. So we flipped the rule. Now the bomb burns the tile you threw it from, and the cost lands on the person who chose it.
Killed because nothing you did mattered
The opposite failure exists too: a game where no choice changes anything. We tried a game where both players tilt the floor under a rolling ball. It produced the cleanest kill number we have ever measured. Every strategy beat every other strategy exactly 50% of the time. Exactly. The game looked alive, but it was a coin flip with nice graphics. Sometimes this failure hides inside a single rule. Our memory game Sleight once had a NOTE button, which let you save a fact for later. The math found the best rate to use it was 0.00 — zero, for every player, even the one losing. A button nobody should ever press is not a game piece. It is furniture. We cut it.
Killed by two humans, after passing every test
The strangest grave belongs to an auction game. It passed seven rounds of measurement. The numbers were healthy: going first was worth only 49.3%, and clever bidding really won. Then two real people played it, and it died in minutes. Not because it was unfair — because it was invisible. All the action happened inside numbers, and nothing on the screen ever visibly moved. That death gave us a rule that now comes before all the math: if a rule cannot be seen as movement on the screen, we do not build it. Computers test the numbers. Humans give the verdict.
What the graveyard bought
Three shapes keep passing the tests, and every fast game we ship is built on one of them. First, give both players one thing to chase — nobody dodges a puck, everybody hunts it. Second, cut the space into tiles or lanes, so good timing beats fast fingers. Third, make every disaster self-made: the ring edge in Sumo Smash, the walls in Trail Duel, the burnt floor in Fuse. In our games you are never simply killed. You are cornered into doing it to yourself. The dead ideas are why we trust these shapes — each one is a law written by a corpse. And the test files stay in our code forever, with names and dates. That way nobody, including us, digs up a dead idea in two years and tries it again.
